Voice Authentication Model for One-time Password Using Deep Learning Models

This paper explores the possibility of implementing a voice authentication system consisting of speech recognition and speaker verication model for the one-time password (OTP) system. The speech recognition model is responsible for classifying user utterances of random OTP digits in Bahasa Indonesia and the speaker verification model is used to verify the identity of the speaker. The long short-term memory network and siamese network with convolutional neural networks are employed as the model, where they aim to recognize and verify human voices represented by MFCC feature vectors. From the experiments, it is found that the validation accuracy of the speech recognition model is reliable, yet the speaker verication model cannot achieve satisfactory result.